Design And Implementation Of SGX-based Face Recognition Privacy Protection Scheme

In recent years,with the development of science and technology,the range of human individual activities has increased and the speed of them has accelerated.The entire society is increasingly demanding safety precautions.Face recognition,as a biometric technology,has high safety for human face diversity,and has achieved initial results as a cross-disciplinary research hotspot.It has been widely used in the field of security prevention.At the same time,most face recognition security applications involve the transmission of huge face images and video data.In the era of the development of the Internet,information technology has penetrated into every aspect of people's daily life.People pay more attention to the growing privacy and security issues than ever before.Face image,as one of the most basic biometric of people,faced with particularly serious privacy security issues.Intel SGX(Intel Software Guard Extensions)technology is a set of Intel CPU-based instructions.It provides users with a secure execution environment secured by hardware.For the problem of privacy protection of image data in face recognition process,we design a scheme that combines the security isolation environment provided by SGX and image encryption technology.In this scheme,there is a client-side and a server-side.The client is responsible for key generation and image data encryption,and the server creates security isolation to perform face recognition model training and face matching.At the first,the user encrypts the face database and transmits it to the server together with the key.The server decrypts the data and trains the face recognition model under the safe isolation.We choose Eigenface,LBP,and fishface which are classic algorithms in face recognition to train model,and the model will be sealed in the cloud.In the process of face recognition,the client transmits the encrypted data to the server.The server reads the model under the secure execution environment,decrypts the data,and performs face matching,at last,the server returns the result to the client.Throughout the entire process,the user's image data is always in an encrypted state in the external environment,and the plaintext state is always in a trusted and safe isolated environment supported by the Intel SGX,which ensuring data security to some extent in the process of face recognition.Finally,using OpenCV,Intel SGX instruction set,this paper implements the proposed scheme and realizes the function test.At the same time,we analyze the execution efficiency and security of the scheme.Our test shows that the system function is stable and the performance is good,which verifies its feasibility.